Bloom Energy is the rarest thing in the AI-power trade: a company where the fundamentals arrived before the story got tired. Most of this sector sells a narrative and dilutes shareholders to fund it. Bloom just reported $1.065 billion of quarterly revenue, up 165% year on year, and $198.9 million of GAAP net income against a $42.2 million loss in the same quarter last year, per its Q2 2026 10-Q. The stock closed at $237.16 on 12 August, up 12.29% on the day and up 140% year to date, per stockanalysis.com. Our scenarios put a bull case at $360 (+52%), a base case at $255 (+7.5%) and a bear case at $130 (−45%).
Here is the number that separates Bloom from almost everything else in this theme. Its weighted-average basic share count rose from 232.5 million to 287.3 million over the year, which is 23.6% dilution – the kind of issuance that normally destroys per-share value. It did not. Revenue per share went from $1.73 to $3.71, a 115% increase. Compare that with the pattern we found at BigBear.ai, where 13% revenue growth against 49% dilution drove revenue per share down 24%. Same sector narrative, opposite arithmetic. Bloom outran its own share issuance by a factor of five, and that single fact is what the bull case rests on.
Key facts
- $237.16 – BE close, 12 August 2026, +12.29% on the day; +140.3% year to date – stockanalysis.com
- $1.065bn – Q2 2026 revenue, up 165.5% from $401.2m – Bloom Energy Q2 10-Q
- $198.9m – Q2 GAAP net income, against a $42.2m loss a year earlier – Bloom Energy Q2 10-Q
- 33.4% – Q2 gross margin, up from 26.7% – FinanceFeeds calculation from the 10-Q
- +115% – growth in revenue per share, despite 23.6% dilution – FinanceFeeds calculation
- $3.9-4.2bn – raised full-year 2026 revenue guidance, with EPS guided to $2.55-2.85
- ~250 MW – onsite power across roughly two dozen AI infrastructure customers after the MiTAC deal
- −32.5% – distance below the 52-week high of about $346 – FinanceFeeds calculation
What the quarter actually said
Revenue of $1.065 billion against $401.2 million a year earlier is not a beat, it is a different company. Half-year revenue reached $1.816 billion versus $727.3 million, up 149.8%. Gross profit tripled to $355.6 million from $107.1 million, taking gross margin from 26.7% to 33.4% – and unlike most hardware businesses scaling fast, Bloom expanded margin while tripling volume rather than buying share with price.
The profitability line is the one that should change how the stock is valued. GAAP net income of $198.9 million in a single quarter, and $272.5 million across the half, ends the argument about whether this is a pre-revenue story. It is not. Reported adjusted EPS came in at $0.78 against roughly $0.41 expected. Management then raised full-year guidance to $3.9-4.2 billion of revenue and $2.55-2.85 of EPS.
One caveat worth stating plainly: a jump of that size in GAAP net income on a business of this scale usually carries non-operating help, and investors should read the tax and other-income lines in the filing before annualising $198.9 million. The revenue, gross margin and cash generation are the durable parts of the quarter. The headline net income figure deserves more scrutiny than it has received.
Why fuel cells won this particular race
The mechanism is timing, not technology. Hyperscalers have capital and chips; what they do not have is grid interconnection. High-voltage transmission takes years to permit and build, and an AI data centre that cannot get power cannot bill anyone. Bloom’s solid-oxide fuel cells run on natural gas, hydrogen or biogas without combustion and get installed behind the meter, on the customer’s site, in months rather than years.
That is why the MiTAC agreement matters more than its size suggests. Adding an islanded fuel-cell microgrid at a Fremont AI server plant took Bloom to roughly two dozen AI infrastructure customers and about 250 MW of onsite generation. Islanded means the site does not depend on the utility at all. For a hyperscaler weighing a two-year interconnection queue against a six-month install, that is the entire decision.
It also explains why Bloom rallied while the grid-scale nuclear names did not. Constellation and Vistra sell electricity that still has to travel; NuScale sells reactors that do not exist yet. Bloom sells the only thing available on the timeline the AI buildout actually runs on. The same interconnection bottleneck is what makes orbital and off-grid compute concepts interesting, a theme we covered when SpaceX partnered with Nvidia on orbital data centres.
Ranking the options on the axis that actually governs the buying decision makes the gap obvious:
| Route to power | Realistic lead time | What it needs first |
|---|---|---|
| Behind-the-meter fuel cells (Bloom) | Months | Gas supply and a site permit |
| Grid interconnection (utility supply) | Multiple years | Queue position and transmission build |
| Existing nuclear via PPA (Constellation) | Contract-speed, but capacity-capped | Spare output at an operating reactor |
| Small modular reactors (NuScale, Oklo) | Late this decade at the earliest | Regulatory approval and a first build |
Read that table as a competitive moat with a shelf life. Bloom’s advantage is not that fuel cells are the cheapest way to make electricity – on a pure cost-per-megawatt-hour basis they are not, and gas feedstock exposes the customer to commodity risk. The advantage is that the alternative is waiting. As long as compute demand outruns transmission capacity, buyers will pay a premium for a system that is generating inside a year. That premium is the margin, and it narrows the moment the queue clears.
The bull case: $360
The bull case needs no change in narrative, only continuation of the operating trend. Take the top end of guidance at $4.2 billion for 2026, assume 2027 revenue of about $5.8 billion as the AI customer base compounds, and apply 18 times sales on 287.3 million basic shares. That produces roughly $360, or 52% above spot.
Three things support it. Margin is still expanding, so incremental revenue converts better than the last dollar did. The customer list is concentrating in AI infrastructure, which is the least price-sensitive buyer in the energy market – these customers are optimising for time-to-power, not cents per kilowatt-hour. And Mizuho’s upgrade to Outperform with a $242 target, citing shipment growth, margin expansion and roughly $27 billion of financing capacity, shows sell-side models still catching up to the reported numbers rather than running ahead of them.
The stock traded at $346 as recently as June, so the bull case does not require a multiple the market has never paid. It requires the market to pay again what it paid eight weeks ago.
The bear case: $130
The bear case is that Bloom is a cyclical hardware company being priced as a software compounder.
At $237.16 and 287.3 million basic shares, the market capitalisation is about $68.1 billion. Against the guidance midpoint of $4.05 billion, that is 16.8 times sales, or 87.8 times the midpoint of EPS guidance. On the diluted count of 323.3 million the same figures are roughly 18.9 times sales. Those are software multiples on a business that manufactures, ships and installs steel boxes.
Hardware businesses have cyclical order books. Roughly two dozen AI customers is a real achievement and also a concentration risk: losing two or three of them, or seeing one hyperscaler pause a buildout, would show up immediately in a quarter. The company has already demonstrated how violently the market re-rates it – the stock fell about a third from its June high before this bounce, and it is still 21.8% below where it traded three months ago even after Wednesday’s 12% jump.
There is also a feedstock question the market has been generous about. Bloom’s systems run predominantly on natural gas, so the customer’s power cost floats with commodity prices and the installation carries a carbon footprint that a nuclear power purchase agreement does not. Buyers with hard net-zero commitments have to reconcile that, and any tightening of procurement standards at a large hyperscaler would narrow the addressable list. It has not happened yet, but it is the risk that would hit the growth rate rather than the multiple, and it is the one least reflected in the current price.
The bear case applies 8 times sales to $4.6 billion of 2027 revenue, which produces roughly $130, about 45% below spot. Note what that scenario assumes: revenue still grows. It is not a demand-collapse case. It is a multiple-compression case, and multiple compression is exactly what happens to hardware companies when the growth rate decelerates from 165% to something merely good.
The base case: $255, and the honest read
Our base case takes 2027 revenue of about $5.4 billion at 13.5 times sales, giving $255, roughly 7.5% above spot. That is close enough to the current price to call the stock fairly valued on central assumptions, which is the honest conclusion.
The asymmetry is in the tails and it is unusually wide: a 52% upside case against a 45% downside case, on a stock with 136% annualised volatility over the past 30 days – the highest of any major energy name we measured. Bloom is not a position to size like a utility. It is priced as an option on the AI buildout continuing at its current pace, and it should be sized accordingly.
For contrast on how differently the market treats an unprofitable version of the same theme, see our analysis of Oklo’s bull and bear cases, and for the capital-intensity problem underneath the whole AI trade, our look at Nebius Q2 and its $5.66bn capex quarter.
What moves the number next
Customer concentration disclosure. Roughly two dozen AI customers on 250 MW means the average deal is large. Any disclosure that names a customer above 10% of revenue reframes the risk profile immediately.
The Q3 margin line. Gross margin at 33.4% while volume tripled is the single most impressive number in the quarter. If it holds through Q3, the software-multiple argument gets much stronger. If it slips back toward the high twenties, the bear case activates without revenue missing at all.
Interconnection reform. Bloom’s advantage is regulatory latency. Any meaningful acceleration in grid interconnection timelines in the US shortens the window in which behind-the-meter generation commands a premium.
Whether the financing capacity gets used. The roughly $27 billion of financing capacity cited in Mizuho’s upgrade is the quiet variable. Deployed as customer financing it accelerates installations without diluting holders further, which is the bull path. Funded instead by issuing equity at these levels, it turns a company that has been outrunning its share count into one that is chasing it – and the BigBear.ai comparison stops being a contrast and becomes a warning.
Our base expectation is that Bloom holds guidance, that the market keeps paying a premium for time-to-power, and that the stock spends the next two quarters between $200 and $300 while investors work out which multiple applies to a profitable hardware company growing triple digits. The bull case needs margin to hold; the bear case needs only deceleration.
This analysis is for information only and is not investment advice. Scenario prices are FinanceFeeds estimates derived from enterprise-value and price-to-sales multiples on 2027 revenue assumptions, and are not price predictions or recommendations. Do your own research.
Frequently asked questions
What is the BE stock forecast for 2026?
Our scenarios put Bloom Energy’s bull case at $360, base case at $255 and bear case at $130, against a spot price of $237.16 on 12 August 2026. The range is wide because the stock carries 136% annualised volatility and its valuation depends heavily on whether the market applies a hardware or a software multiple.
Is Bloom Energy profitable?
Yes. Bloom reported GAAP net income of $198.9 million in Q2 2026 and $272.5 million across the first half, against losses of $42.2 million and $65.6 million in the comparable 2025 periods. Reported adjusted EPS was $0.78 versus roughly $0.41 expected. Investors should still read the non-operating lines before annualising that figure.
Why did Bloom Energy stock rise so much in 2026?
Revenue grew 165% year on year in Q2 to $1.065 billion as AI data centre operators bought behind-the-meter power. Grid interconnection can take years, while Bloom’s fuel cells install in months, so hyperscalers facing a power bottleneck pay for speed. The stock is up 140.3% year to date.
Is Bloom Energy diluting shareholders?
Yes, but productively. Weighted-average basic shares rose 23.6% year on year, from 232.5 million to 287.3 million. Revenue per share still grew 115%, from $1.73 to $3.71, because revenue grew far faster than the share count. That is the opposite of the pattern seen at most small-cap AI names.
Is BE stock expensive at $237?
On the guidance midpoint of $4.05 billion, Bloom trades at roughly 16.8 times sales on basic shares and about 87.8 times the midpoint of EPS guidance. That is a premium multiple for a manufacturer, and it is the core of the bear case: the business is real, but it is priced for the growth rate to persist.
What is the biggest risk to Bloom Energy?
Customer concentration combined with multiple compression. About two dozen AI customers account for roughly 250 MW, so a single hyperscaler pausing a buildout would be visible in one quarter. Because the stock trades on a software-like multiple, deceleration alone – without any revenue decline – is enough to trigger a large de-rating.